Research On Disordered Sorting Technology Of Industrial Robot Based On 3D Target Recognition

In recent years,a large,number of industrial robots have been applied to assembly and manufacturing,which has led to an increase in the automation level of the company's production process.This also makes the company's production mode shift from the traditional manpower-led to the robot-oriented.This mode of change greatly improves the production efficiency of the product,reduces the production cost,and reduces some of the harsh production environment.The harm caused by the worker's body.In the process of robot operation,how to accurately capture the parts or goods piled up randomly is a hot issue in current research.This process mainly involves the acquisition of pose information of target components.This paper designs an industrial robot disorder sorting system based on 3D target recognition to carry out experimental research on the above problems,which has certain research significance and practical value.The main research work of this paper is as follows:(1)Designed a disorderly sorting platform solution.The programming method of the robot in this paper is different from the traditional method of teaching and offline programming.It is to use the camera to visually detect the sorted object to obtain the pose information of the target object.Therefore,it must be based on the environment and the specific object to be sorted.The infonnation is selected for the hardware of the sorting system,and the process and scheme of the whole system are designed.The design system should meet the centimeter-level grabbing requirements.(2)The calibration problem of the robot vision system composed of Kinect camera and ABB 1200 robot is studied.This paper determines the connection between the Kinect camera and the robot as the eye-mounting method.The MATLAB is used to calibrate the Kinect camera itself to obtain the error of each calibration image,the overall average error and the camera's own parameters.After that,the robot and The Eye-to-hand system consisting of the camera performs hand-eye calibration to obtain the positional relationship between the camera coordinate system and the robot base coordinate system.(3)The pose estimation problem of multiple target obj ects randomly stacked is studied.In this paper,the point cloud information collected by the camera is preprocessed,the RANSAC algorithm is used to fit and remove the plane of the target obj ects,and the template is extracted by the same method.Then,the clustering algorithm is used to target multiple targets.The obj ect is segmented,and then the OUR-CVFH feature is used to classify the target object according to various different perspectives.Then the SHOT feature is used to initially match the template with the target obj ect,and then the ICP algorithm is used for further point cloud registration.Finally,combined with the results of the hand-eye calibration,the pose of the target object can be estimated.(4)The experimental platform of the whole disordered sorting system was completed,and the multiple-order disordered experiment verification of multiple target obj ects was carried out,and the experimental results were analyzed.The result of the target pose estimation and the actual position error is about 5mm,which verifies the feasibility of the system scheme and method,and can complete the capture and placement of multiple randomly stacked target objects,so that the robot can perform centimeter-level without Sorting tasks to meet the design requirements of the system.